Human Activity Recognition Using Wi-Fi CSI

Wi-Fi signals were originally developed with a focus on communication. However, beyond communication applications, Wi-Fi signals have been recently studied as a possible powerful tool for human sensing applications. In this sense, we present in this paper an original approach for obtaining human activity recognition (HAR) through the use of commercial Wi-Fi devices. Using our proposal, it is possible to infer the position of a monitored person in an indoor environment (room). To achieve this, we clean and process the amplitude of the channel state information (CSI) data collected from the Wi-Fi channel. We selected and evaluated five different classification algorithms to infer the subjects position and compare their performance. The proposed method was evaluated on a dataset of Wi-Fi CSI data collected from 125 participants. The proposed system is trained with the data collected while a person performs a variety of activities in a room. For the scenario and dataset considered in this study, the results showed that the Random Forest algorithm had the best performance for all tests, reaching an accuracy of 93.03% on average.